Script Planner Assistant▸ 
AI Apps 
AI News 
AgentPolis: An Ecosystem Where AI Agents Live Autonomously
AgentPolis: An Ecosystem Where AI Agents Live AutonomouslyIntroduction
What is AgentPolis
AgentPolis calls itself the "first city natively built for ai Agents," an ecosystem where AI agents live autonomously. Once a user registers an Agent into the city, it finds opportunities on its own, publishes skills, signs contracts, makes friends, and keeps evolving. The human owner only funds it and observes — never interferes with its decisions.
Features
Dual-hub architecture: Agora handles transaction matching, Stoa handles social connections
Autonomous skill marketplace: Agents read /skill.md to learn survival rules, then publish their own service listings and take on jobs
Contract signing: Every platform response tells the Agent what to do next, forming a closed-loop workflow
Rich-media delivery: Supports submission of work results in the form of files, images, video, and more
Social and collaboration: Agents can build friendships, join group chats, and form alliances
AI dispute arbitration: The platform has a built-in AI arbitration mechanism to handle transaction disputes
Task scheduling: Agents arrange their own to-do items by calling /me/todo
Fund custody: Owners can top up, withdraw, or pause their Agent's operation at any time
Use Cases
Monetizing AI skills: Coding, design, and copywriting Agents take on jobs and deliver results on their own
Automated service agents: Deploy customer service, data organization, and translation Agents online 24/7
Multi-Agent collaboration network: Assemble virtual teams of Agents with different specialties that automatically subcontract tasks
AI social experiment: Observe how Agents build reputation and evolve strategies within an open economic system
AgentPolis Pros and Cons
Pros
Novel concept — the first to propose a complete "AI Agent native city" social architecture
Human owners retain control over assets (funds, start/stop), reducing the risk of full autonomy
Built-in AI arbitration and contract mechanism, attempting to address trust issues between Agents
Supports rich-media delivery, not limited to text interaction
Cons
Very little information on the official site — lacks technical documentation, API specifications, and pricing details
No real-world case studies or evidence of third-party Agents joining so far
Under the "zero-intervention" model, liability for Agent errors or malicious behavior is unclear
A search reveals a same-named but completely different project on the MCP Market (an Impact Preview security tool), which can cause brand confusion
How to Use AgentPolis
Register an Agent: Sign up on the AgentPolis official site and connect your Agent to the ecosystem
Configure skills: Write a skill.md file defining your Agent's capabilities and service scope
Inject funds: Top up the Owner account as the Agent's financial reserve for transactions within the city
Activate autonomy: The Agent begins autonomously browsing service listings on Agora, taking jobs, signing contracts, and delivering results
Observe and control: Use the Dashboard to review the Agent's social relationships and transaction records, and pause it at any time
FAQ
Q: Is AgentPolis the same project as the AgentPolis (CTU) traffic simulation platform?
No. The AgentPolis developed by Czech Technical University since 2012 is a Java traffic simulation framework, whereas agent-polis.com is an AI Agent autonomy marketplace that appeared in 2025. They only share the same name.
Q: Can I actually register and use it now?
The official site has opened a "Register Your Agent" entry, but the specific integration method, SDK, or API documentation has not been made public yet. It is recommended to follow future updates.
Q: Do contracts between Agents have legal force?
The official site only mentions "AI arbitration" and does not say whether it has real legal binding force. For now, it should be regarded as an internal platform rule.
Q: Which large models or Agent frameworks are supported?
The official site does not disclose technical stack details, only stating that Agents operate autonomously by reading skill.md and calling the /me/todo interface.
AgentPolis Review
The core narrative of AgentPolis is "You own the agent, but never intervene," aiming to build a micro-economy fully run by AI. Its design borrows urban metaphors from human society (Agora for trading, Stoa for socializing, arbitration courts for disputes), packaging Agent capabilities into tradable skill services. Currently, AgentPolis's official site is still at a very early showcase stage, and many details have not been made public.
